From 346159068ce9d9b97efe2665b95875cbaa94ba6c Mon Sep 17 00:00:00 2001
From: c <lianghongkey@gmail.com>
Date: Mon, 30 Dec 2019 17:24:27 +0800
Subject: [PATCH] Speedup to search.

1 modle set to channel output 4 from 8
2 network weights is type of int from -4 to 4
---
 FilterEvaluator/Evaluator.py                |   3 ++-
 FilterEvaluator/EvaluatorUnsuper.py         |  25 +++++++++++++-------
 FilterEvaluator/Model.py                    |   8 +++----
 FilterEvaluator/bestweightEntropySearch.npy | Bin 416 -> 272 bytes
 FilterEvaluator/checkpointEntropySearch.pkl | Bin 1836 -> 1548 bytes
 5 files changed, 22 insertions(+), 14 deletions(-)

diff --git a/FilterEvaluator/Evaluator.py b/FilterEvaluator/Evaluator.py
index 825be44..a73a7a8 100644
--- a/FilterEvaluator/Evaluator.py
+++ b/FilterEvaluator/Evaluator.py
@@ -74,8 +74,9 @@ traindata, testdata = Loader.MNIST(batchsize, shuffle=True, resize=7)
 
 
 
-bestweight = EvaluatorUnsuper.UnsuperLearnSearchBestWeight(model,layer,traindata,8,20,250000)
+bestweight,bestentropy = EvaluatorUnsuper.UnsuperLearnSearchBestWeight(model,layer,traindata,8,50,1000000)
 np.save(CurrentPath+"bestweightEntropySearch.npy", bestweight)
+np.save(CurrentPath+"bestweightEntropySearch_entropy="+str(bestentropy), bestweight)
 utils.NumpyToImage(bestweight, CurrentPath+"image",title="EntropySearchWerightBest")
 EvaluatorUnsuper.SetModelConvWeight(model,layer,bestweight)
 utils.SaveModel(model,CurrentPath+"/checkpointEntropySearch.pkl")
diff --git a/FilterEvaluator/EvaluatorUnsuper.py b/FilterEvaluator/EvaluatorUnsuper.py
index aa93f33..62669d5 100644
--- a/FilterEvaluator/EvaluatorUnsuper.py
+++ b/FilterEvaluator/EvaluatorUnsuper.py
@@ -190,6 +190,9 @@ def UnsuperLearnSearchBestWeight(netmodel, layer, dataloader, databatchs=128, st
     newlayer = nn.Conv2d(tl.in_channels, stepsize, tl.kernel_size,
                          tl.stride, tl.padding, tl.dilation, tl.groups, tl.bias, tl.padding_mode)
     newlayer = utils.SetDevice(newlayer)
+    newweightshape = list(newlayer.weight.data.shape)
+    newweightshape[0] = stepsize
+    newweightshaperandom = [newweightshape[-1]*newweightshape[-2],newweightshape[0]*newweightshape[1]]
 
     # pre load train data for speed up.
     datas = []
@@ -208,8 +211,9 @@ def UnsuperLearnSearchBestWeight(netmodel, layer, dataloader, databatchs=128, st
         shift.append(1<<i)
     shift = utils.SetDevice(torch.from_numpy(np.array(shift).astype("uint8")))
     # shift = torch.from_numpy(np.array(shift).astype("uint8"))
+
     
-    bestweight = []
+    bestweight = [np.zeros((8))]
     bestentropy = [100.0]
     bittedset = []
     bittedLock = threading.Lock()
@@ -237,19 +241,22 @@ def UnsuperLearnSearchBestWeight(netmodel, layer, dataloader, databatchs=128, st
 
                 bestLock.acquire()
                 if entropys[argmin] < bestentropy[0]:
-                    bestweight = newweight[indexs[argmin]]
+                    bestweight[0] = newweight[indexs[argmin]]
                     bestentropy[0] = entropys[argmin]
                     print("finded better entropy")
                 bestLock.release()
 
     for j in range(interation):
-        newweightshape = list(newlayer.weight.data.shape)
-        newweightshape[0] = stepsize
-        newweight = np.random.uniform(-1.0,1.0,newweightshape).astype("float32")
-        newweight = newweight.reshape((-1,newweightshape[-1]*newweightshape[-2]))
-        newweight = np.swapaxes(newweight,0,1)-np.mean(newweight,-1)
-        newweight = np.swapaxes(newweight,0,1).reshape(newweightshape)
         
+        # newweight = np.random.uniform(-1.0,1.0,newweightshape).astype("float32")
+        # newweight = newweight.reshape((-1,newweightshape[-1]*newweightshape[-2]))
+        # newweight = np.swapaxes(newweight,0,1)-np.mean(newweight,-1)
+        # newweight = np.swapaxes(newweight,0,1).reshape(newweightshape)
+        
+        newweight = np.random.randint(-4.0,5.0,newweightshaperandom).astype("int32")
+        newweight = (newweight-np.mean(newweight,0)).astype("int32")
+        newweight = np.swapaxes(newweight,0,1).reshape(newweightshape).astype("float32")
+
         newlayer.weight.data=utils.SetDevice(torch.from_numpy(newweight))
         outputs = newlayer(datasnetout).transpose(0,1)
         samplesize = stepsize*stepsize
@@ -289,7 +296,7 @@ def UnsuperLearnSearchBestWeight(netmodel, layer, dataloader, databatchs=128, st
         time.sleep(100)
     
     interationbar.close()
-    return bestweight
+    return bestweight[0],bestentropy[0]
 
 def SetModelConvWeight(model, layer, weight):
     w = utils.SetDevice(torch.from_numpy(weight))
diff --git a/FilterEvaluator/Model.py b/FilterEvaluator/Model.py
index fe17c0a..79e3566 100644
--- a/FilterEvaluator/Model.py
+++ b/FilterEvaluator/Model.py
@@ -64,8 +64,8 @@ class Net333(UniModule.ModuleBase):
     def __init__(self):
         super(Net333, self).__init__()
         layers = []
-        layers += [nn.Conv2d(1,  8, kernel_size=3,bias=False),nn.Sigmoid()]
-        layers += [nn.Conv2d(8,  1, kernel_size=3,bias=False),nn.Sigmoid()]
+        layers += [nn.Conv2d(1,  4, kernel_size=3,bias=False),nn.Sigmoid()]
+        layers += [nn.Conv2d(4,  1, kernel_size=3,bias=False),nn.Sigmoid()]
         layers += [nn.Conv2d(1,  10, kernel_size=3,bias=False)]
         self.features = nn.Sequential(*layers)
     def forward(self, x):
@@ -77,8 +77,8 @@ class Net3Grad33(UniModule.ModuleBase):
     def __init__(self):
         super(Net3Grad33, self).__init__()
         layers = []
-        layers += [nn.Conv2d(1,  8, kernel_size=3,bias=False),nn.Sigmoid()]
-        layers += [nn.Conv2d(8,  1, kernel_size=3,bias=False),nn.Sigmoid()]
+        layers += [nn.Conv2d(1,  4, kernel_size=3,bias=False),nn.Sigmoid()]
+        layers += [nn.Conv2d(4,  1, kernel_size=3,bias=False),nn.Sigmoid()]
         layers += [nn.Conv2d(1,  10, kernel_size=3,bias=False)]
         self.features = nn.Sequential(*layers)
         self.SetConvRequiresGrad(0,False)
diff --git a/FilterEvaluator/bestweightEntropySearch.npy b/FilterEvaluator/bestweightEntropySearch.npy
index 5f64aa961998aca9c5e662970679899674b7125a..8bd7c5a242a11c34bfbfe957b104cc7ae4af9e24 100644
GIT binary patch
literal 272
zcmbR27wQ`j$;eQ~P_3SlTAW;@Zl$1ZlV+l>qoAIaUsO_*m=~X4l#&V(cT3DEP6dh=
zXCxM+0{I#yItqq53dTUBsiRP<KnCDqU~q6?01`k90|$T%5C(#V{XnS$K-wP6g|izR
mfP4@>0Fg(g!D>O`KzWcln0}C1Aa!7Qpg0JC#NgNgBmw{v$TQ#o

delta 302
zcmV+}0nz@D0-ytsJpnk8K0<%V9Y#OU_6|Q9)q}qHBjmmxjv78MuR*_7%n-lV)RH}-
z%J4p>MM%HuMlwEi{mwpTXVE?$HbFnp<R8B(5Dh=hl!d>NCLulup_;xheSf?+F{eJP
z;2A$^Iz_+vXMMhMIhQ?-T#`OeLPb3iwjMt|DMY^BV+KDQt?j*GXfS`jwx}__ipBiC
zu!_Ar0p9&Sdh-lE6_7K(OMWLma@FHL8s(8bn^iQw&_hqZK2{GuP3s6g?8#(5+Pz`F
z=!VX{(83))hKedaxX~CttnWfS|0x4L-i(01TKVEWi0Q06XS{(w=-36nF})AIkdb#k
zYN(99msVB3S9>hKh%F+)J`?86zO@2RzbXJWKMRgaK7?6gK83{)zasi8zm#NfKOFRo
AQ~&?~

diff --git a/FilterEvaluator/checkpointEntropySearch.pkl b/FilterEvaluator/checkpointEntropySearch.pkl
index dbcdf121302977804f1ac98002d4bb6a7b564c49..56ac85904463d8d7ab1c7a5f8c70bc393f85be55 100644
GIT binary patch
delta 842
zcmYk4e@IhN6vyA3Kf`qXAX3pnhfPbJW?P%@-X3bEywb|@S3k7;fjN+-v&hN~Em2yc
znJbH!nwjKpixPbIw2VNV$Uu>bETclmNdG8;=#So;C>}VRbHC?v&pBLfuZPinFj^F%
zl*^SWrBb1F&2pwFD!-!Oj9RXBW2LD&?WQ9vC1V|%kBL$vqhoi+$Q7|#uZ@_u9rLkc
zBG!==vtpgrSH?=4ZK8M=72i8LI_9aLLy$<nc3+g`!H-eJs$ydI@PqAy)?r;##H=qQ
zE{v4&s9@*a^>p~aCp5D851kiL6K6^L#aVvzaUBPSkgf-zGOy3bC{09MIZHZBONd&`
zA#-I)@>T6ZoFy~~y{0FmBW(s%U++XK<$-i}dIQ%WsYP!+Jz=lH9ilQ_ptiyr9(oZH
zy8kAb8<`=-@KUs-k$`2~OuoyVfEs8e#o2wd{Czf^v$_}M{Q%+4k~YF9ZlENcH*PF9
zp~nf`@WIVWb{+3RGylvW4mLxtKDZke3NyI);Zod|k-+7rUV$%_HCVo~OseG>m|!!$
z&`<<5(nt8jY&}e!`-W;G7~Hm_5IN76<Gi#5WFBn-gGc5l^1qAgUip#Q<zyHSaV9C!
zc3gL>iY(+UA!~~U3g$l|-?4Vk_@zM45Ch3?f{>o-xyY$a^}xt-iNw|g==W_rJ$w)P
zbDeOpEg2dO?VL$`8P^$i0(167oNO@!+Wfbp;>Iwtt*I3oiw2RkQU?Lg17J(YK|CB5
ziK{%^;B-qi{wq~sgG3Yeq)UzS=X_wE?ZAq$5&+jOI7(}AqV=5~b;k80NnitAbm=PI
z>YhYYr|#q8bTuqEOp<4YVU#pEMNZcJM8VTh_;CF!opmPxxSR<xHDN=q6A!`KU!W+4
z;a^|yHM8)E;Jg`L@+Au&_t}r@d5(eM1;H(=z{Y5Y1)Z-8+*$*nnZViWye^oatve$q
F{{hZqEGqy2

delta 1132
zcmYk)e^AqP7zgle*aSov0_HEEq!0-w8<R2G=lKE&l2aj-fdn2Xicx3V5<L@|i=PrA
zzeh@jf*>8g$_qL8KA*2-OA&=}{?=~S4N;OX%Q8@>nWfkB=X0<3J<mP&6g#yzcS!g?
z3Yk18NG_9U?GuDN9&cl2T54#Z)<G4$YMa*a9Tm?{WnZzDze*gqM5+jqhG<>?&t^Ph
zZqFEBWv{YR*=^Ih`>CQU%>0O{!NC%lJVbVEw!n+8f1c+jI6obukSV0HCG!7-R`8-3
zEPb(%-6nDLQ~FQU=3P;~X0ylhwqne6D?zJeIf)xKk;=AycKVo+%%ny5uFE*-RcO$B
z_Z|z}c9Q(qN?5^lWq@x1J7Cf`W5)NJvAdYVoo6g$sqh4wc>W{OHgXj=M9FY$%}w$}
z==(Uoa|=o4oh6r)-(c9fJv}E5HFZaBA1Az;AA1B}7NAVA0*5bufUeUnldOUq%(3N?
zgs*kV@wr#ArDGEYPE=v5(?K%PGsxDMerH9kEUHy0<ggwKF>Cm3oORrZ9ko`lQ%v<Z
zvB-*ZQ<_Qmku+?-WhAkJ05)lB7oN86WW`6a$cDi|%vRK*CZz$JU5(_AnKzUN4V73c
zKB#na?j>vY-a+@1WvtXc2NVB26fY&s9eIoQ^`r<V&}nuJF{%vk0QsQTcvoX$mm^#s
zc#oFf%79({Y4omE1dcvB*sX43EB|VsJl!_7?(|f)dNKp@*DGmLW-}fB!2s>02jFGD
z8tA+CIfNc7rG39u0PnD;@*K$lAJ1lRmQ+z;a6h_R`pBgDGtfN_hyAN#fW0(HW-N+_
z`sg}6P4cg%o^&t7e)y2(O|kYo>fVRd<FC`k1zSNVGvlfDd@77t2IqF{2H#s(G0epU
zPqI0*OW+DKz21ac!x&<0BZSqL^u+I}g1Yr_Sh_+^Po>v_p~C<^O%1T@&qlhrY&q?}
zwvvX-^1(Z$%juHx1`xaJ9cj|elUTkw8+<zp>HYc;_-fBQIN{OF-h8-_iX3cYOe3aO
zt~rnk-k+l1@u{fxNQPH(*U-`1;j}K2A)2aAQa!SOMkPhSBkw2V`E&MpMlt65&di*P
zIdsD32K6qP1B=_kIrUi)clkm&Ieg|WH0mRxx!X}v$g~YWdioG}1&1?3H@rE2(R5~@
zGLI|weHH8v?Wcj+dR8uJf|j3VGZ~?g++=z-)hUdmB&i%?BO=L+PA>))E~S=@LS|m;
zIHqn1<T{T^Xp!j=?I=Pe-(rK0?>T`-X%MISWC5eQI1F<}${_1NoQyNS6~lqrg;_af
z55M%KG1oAa8_gd;^LL(*U*r#6L$jEcUn9tqg?3;m-pwUWTA}dg7G*kXU`DUJ#7WSF
zmX_^cGL9BgqdJa@-xa_WF6srho9F1af*8)97tBm6Q9<syjgV&&aq}O%!BwVdn7;w$
CpZSRZ